Research On Trading-off Between Accuracy And Energy Consumption Of The Auto-correlation Based Step Counting Algorithm

With the development of mobile Internet and wearable devices,step counting algorithms have been widely applied in physical exercise,health management,pedestrian positioning and navigation,and etc.The accuracy of step counting algorithms has significant influences on the service quality of these applications.However,existing studies on the step counting accuracy are mostly realized by experimental methods and lack of theoretical foundations.Besides,in long-term applications,energy consumed by step counting is also ignored.To address the above issues,we will carry out the following researches.Firstly,a thorough statistical analysis and a large number of experiments are conducted with the assistance of a high-precision and professional equipment,namely Xsens MTw,which is used as an error calibration tool for the measurement errors of the inertial sensors embedded in smartphones.Secondly,taking the auto-correlation based step counting algorithm as an example,we utilize the probability theory and Taylor expansions to establish a step counting error model,which reveals that the step counting accuracy is dependent on the measurement errors of inertial sensors,sampling frequency and device placement.In particular,the lower measurement errors of the inertial sensors,the more accurate step counting is;the step counting accuracy will increase and gradually become stable with the increase of the sampling frequency;the step counting is more accurate when the device placement introduces a large amplitude of the movement.After that,abundant experiments are done and validate the established error model.Finally,the constraint relationship between the energy consumption of step counting and the sampling frequency of the inertial sensors is obtained by measuring the energy consumption of step counting in smartphones,and is useful for the trading-off between the step counting accuracy and the resulting energy consumption.Especially,the experiment indicates that,without significantly decreasing the step counting accuracy,the step energy consumption can be saved by around 41 mAh per hour.To sum up,the research outcomes in this thesis provide a solid theory and an experimental basis for the study of step counting algorithms and relevant applications,and create a new perspective in the corresponding fields.